But I thought they were re-usable?
A full stack Starship right now needs 39 engines (33 for Super Heavy, 6 for Starship). SpaceX will definitely lose some to testing (and indeed last I checked they are 100% going to lose all the initial ones, because they're going to do a mock-landing over the ocean before attempting a real landing, so even if it goes perfectly all those are going into the salt water). Once they have SH being recovered, ideally starting launch 2, that will cut the burn rate big time but they'll still probably take a bit to start successfully recovering SS. That's entirely expected and acceptable for MVP and as a worst case for HLS (because Lunar Starship and such aren't coming back to Earth at all, that's Orion's job). SpaceX runs hardware rich, so they will need hundreds just for testing.
Long term though they want 1000+ Starships for Mars and other efforts. There will still be expended Starship profiles, and reuse doesn't mean that cadence won't require having more raw vehicles. Fueling orbital depots quickly requires lots of vehicles and trips, and boil off particularly early on means just depending on a few isn't ideal (and that's assuming they have reuse working by then, if they have to just eat the refueling as expended to meet their contract obviously they'll need more). With the current design that's tens of thousands of engines, so at current production they'd be hitting the right numbers in about 22 years which seems in the ballpark of when such an effort might actually get going.
At any rate mass production is very important for their plans and how they operate. It's let them iterate very quickly, iron out more kinks, be much more aggressive about testing which in turn feeds back into doing each better, and pushes down the cost per engine dramatically. If you go to the Ars comments first page you'll see pictures comparing Raptor 1 and Raptor 2 and can just visually see the massive simplification (though note it's a little bit unfair because Raptor 1 was much more heavily instrumented, but there is far more stuff removed/improved then just that).
Are they going to pick them up or let them sink?
https://www.teslarati.com/spacex-elon-musk-starship-orbital-...
(Although I guess you could refly the same tanker a few times)
If you were really going to Mars you would need all the Starships you could get, even operations around the Earth and cislunar space could keep a substantial fleet busy if the cost per pound is low enough to encourage new applications. You will also need to replace engines at some rate.
Inventory->Hardware->Raptor2 -= 33;If it manages to "land" successfully just above the water, then they'll go for landings back at the pad.